By "enter" do you mean put the data into SPSS?
A "check all that apply" type question is usually entered as a multiple
dichotomy set.
One common problem with survey instruments that are not carefully
designed is that the response categories are not exhaustive, I.e., they
do not distinguish "not answering the question at all" and answering the
question "none of these".
Do you have this problem? If so, while reviewing the raw instruments,
see if there is evidence that the respondent did not answer the question
at all vs having none apply, create a variable called none with a one if
none apply. (By any luck do you have access to the respondents for
follow-up?) Of course, a lot depends on the actual wording of the
question and the context. It may be that Other may have the same
semantic status as NONE is this situation. During your quality
assurance debriefing of respondents what reasons did they give for not
answering at all?
For each of the items and the new variable enter a one, for each item
that is not checked, enter a zero, if you believe the question was not
answered at all enter -1.
value labels S1, O2, P3, Other, None
1 'checked'
0 'not checked'.
-1 'whole question skipped'
-2 'legitimate skip'.
missing values S1, O2, P3, Other, None (lo thru -1).
For the regression try something like this untested syntax.
regression variables = my_dv S1, O2, P3, Other, None
/dependent = my_dv
/method = enter S1, O2, P3, None.
If I recall correctly this will give you the zero order
coefficient(i.e., as if the variable were the only predictor) in the
section called "variables not in the equation.
It also will give you the fit for the whole set of predictors treating
other as redundant.
Art Kendall
Social Research Consultants
OZGU GILL wrote:
> Hi all,
>
> I have a simple survey results entry question in deed:
>
> I need to run a regression for some survey questions. In one of the questions the respondent can choose as many as s/he wishes among 4 choices. Some respondents have only 1 choice some of them didn't answer that question at all and some of them have chosen more than 1 choice even 2 choices. The choices are S1,O2, P3and Other. I'm parrticularly interested in seeing how much S1 or P2 is affecting my dependent variable. How should I enter the responses to include this question in the regression equation. Thanks in advance.
>
> Regards,
>
> Ozgu
>
>
>---------------------------------
>8:00? 8:25? 8:40? Find a flick in no time
> with theYahoo! Search movie showtime shortcut.
>
>
>
>
Art Kendall
Social Research Consultants
Locked
